Market Competition and Efficient Cooperation
Jordi Brandts
Arno Riedl∗
August 26, 2015
This draft is prepared for the website of the Amsterdam conference ‘‘Endogenous
Preferences and the Broader Effects of Competition‘‘ (1-2 September 2015). It
is incomplete and preliminary and should neither be further circulated nor
quoted without permission of the authors.
Abstract
We use laboratory experiments to study the effects of competing in a market under favorable
and unfavorable conditions on cooperation in a subsequent social dilemma game. The issues
we study are part of the broader topic of whether there are behavioral spillovers between
different spheres of social interactions. Market interaction takes place in a continuous double
auction market in which one side of the market obtains the whole surplus. We examine
subsequent cooperation behavior for pairs of market-winners, market-losers and mixed pairs
and study both the cases where interaction in the social dilemma is with others from the
same market, ‘market-partners’, and where it is with others from another market, ‘marketstrangers’, and compare it with benchmark behavior in a stand-alone social dilemma game.
We find that in market-partners competitive market experience has adverse effects on efficient
cooperation on both market-winner and market-loser pairs. In market-strangers pairs of
market-winners manage to cooperate more efficiently. These results indicate that it is not
market experience per se that lowers the ability to cooperate. Having competed for scarce
resources on the same side of the market makes it difficult to overcome the social dilemma and
positive market experience fosters cooperation only for those who did not have to compete
with each other. We also show that differences in cooperation cannot be explained by ex-ante
income differences and also find evidence that market experience affects subjective well-being
and social value orientation.
Keywords: Competition, Cooperation, Experiments
JEL: A13, C92, D30, J50, M50
∗
Jordi Brandts: Institut d’Anlisi Econmica (CSIC) and Barcelona GSE; [email protected]; Arno
Riedl: CESifo, IZA, Netspar, and Maastricht University, Department of Economics (AE1), P.O. Box 616, 6200
MD Maastricht, the Netherlands, [email protected]. The authors thank the Spanish Ministerio
de Economa y Competitividad (Grant: ECO2011-29847-C02-01), the Generalitat de Catalunya (Grant: 2009 SGR
820), the Antoni Serra Ramoneda Research Chair (UAB-Catalunya Caixa) for financial support.
1
Introduction
In economics markets are traditionally studied separately from other spheres of social interaction,
with the focus of analysis typically being on the allocation of resources and without taking into
account possible spillovers effects between market activity and other types of interaction. This
way of proceeding may be seen as a covenient simplifying research strategy. However, it can
lead to important omissions in our understanding of the effects of markets. The question if and
how competitive markets may spill over to other spheres of social and economic life has recently
also been receiving some attention in the public debate. These discussions suggestive, but clean
evidence on the matter is still scarce.
In this paper we use lab experiments to study the casual effect of market experience on
efficiency of voluntary cooperation. That is, we explore whether the experience of participating
in a competitive market leads to efficiency gains or to efficiency losses outside the market, or
whether it has no effects. To study this, we compare behavior in a social dilemma game in
the absence of market interaction to behavior when the social dilemma game it is preceded by
market interaction. Market interaction takes place in a highly competitive continuous double
auction (see Smith, 1962), with a so-called box-design (see Holt et al., 1986). In experiments,
the double auction has been shown to be the institution that best embodies the characteristics of
markets where prices and transactions come about through a decentralized equilibrating process,
that converges to the Walrasian outcome (Davis and Holt, 1993). The box-design involves a long
and a short side of the market. In a stylized way, it represents market conditions in which the
individuals on one side of the market will easily make transactions at favorable prices, whereas
the individuals on the other side of the market will have difficulties to make transactions and
will do so at unfavorable prices, if they transact at all. This allows us to explore the effect of
different market experiences. The social dilemma game we use is a repeated two-person public
goods game in which pairs are fixed through all rounds. Public goods games are a standard tool
for studying cooperation (Chaudhuri, 2011; Kagel and Roth, 2012).
Our study relates to the broader issue of the influence of society on preferences, which is
an under-explored topic in economics. Fehr and Hoff (2011) argue that economists’ traditional
reluctance to explain changes in behavior in terms of preferences was partially motivated by
the difficulty of producing conclusive scientific evidence in favor of the preferences hypothesis
with natural data. Given this difficulty they suggest that the use of experimental data (field,
natural and quasi-experimental) may be a promising strategy for learning more about this issue.
1
Relatedly, van Winden (2012) argues that to understand relationships between people one needs
to take into account the existence of social ties and how they are affected by the context in which
social interactions take place. Bowles and Polania-Reyes (2012) present an extensive survey of
the evidence that documents that social motivations are not separable from other experiences.
In economics, competition is commonly seen as a beneficial force, as for example when more
firms in a market are seen as leading to lower prices and, in a more dynamic view of markets,
when entry of new firms into a market is considered to generate more innovation and cost savings.
It is without doubt that competitive markets can lead to efficient allocations of resources, as
demonstrated theoretically in the First and Second Welfare Theorems (see, e.g., Mas-Colell et al.,
1995) and empirically in experiments with double auction (and other) markets (see Smith, 1962,
for the seminal paper and Davis and Holt, 1993, for a survey).
However, there are efficiency effects of markets conceivable that reach beyond the market
environment itself. An important feature of modern market societies is that the productive assets
including human capital that allow people to earn income are distributed rather unequally. As
a consequence, some peoples skills or assets will face high demand in the market, with many
others trying to transact with them, while those of others are in much lower demand. Depending
on market conditions, some people may have difficulties to trade at all. Moreover, for many
individuals being in a favorable or unfavorable market situation will be constant over their lifetime; often it even carries over between generations.The most prominent case of such situations
are labor markets where a large number of low-skilled workers compete for a limited number of
jobs. The question arises whether such very different market experiences will affect the efficiency
of interaction in spheres where cooperative behavior outside a market setting is necessary to
achieve efficient outcomes. Depending on the results, the analysis of such spillover effects may
ask for a reassessment of the virtues of market competition.
Two contrasting views on the spillover effects of markets on cooperative non-market environments have been proposed by some prominent researchers. Smith (1998) builds on Adam
Smith to postulate that people intuitively know how to behave both in a cooperative and in a
competitive way depending on the context. According to this view, both behaviors grow out of
a universal propensity for social exchange which “finds expression in both personal exchange in
small-group social transactions and in impersonal trade through large-group markets” (Smith,
1998, p.3). Smith sees cooperative and non-cooperative behavior as peacefully coexisting, with
efficiency in impersonal markets being based on competitive behavior, while efficiency in personal
social and economic exchange requires reciprocity.
2
Henrich et al. (2001) go a step further and suggest that market interaction may have positive effects on cooperation. They study behavior in ultimatum games, public good games and
dictator games in 15 small-scale with a variety of economic and cultural conditions and relate
the results to a non-experimental measure of market integration. They find that “the higher the
degree of market integration (...) the greater the level of cooperation in experimental games.”
(Henrich et al., 2001, p.74) The rationale for this relation proposed by these authors is that
“the more frequently people experience market transactions, the more they will also experience
abstract sharing principles concerning behaviors towards strangers (...).” (Henrich et al., 2001,
p.76) This is consistent with the notion of doux commerce as put forward among others by
Montesquieu (1748).
In contrast, Bowles (1998) argues that market participation can transform people: “(...) there
are significant differences in the personality effects on participants in markets (...) for people on
the short side (...) and those on the long side of the market, some of which are simply excluded
from the exchange process, while others fear losing the transactions they have secured” (Bowles,
1998, p.78). Bowles’ concerns can be seen as part of the broader issue whether market exchange
erodes certain moral and civic goods worth preserving. Our focus will be on whether market
experience affects the ability to interact with others to achieve efficiency.1
The diverging views on the potential spillover effects of market participation call for empirical work that can provide insights into their relative merits. If Bowles is right about the
negative spillover effects of market participation on cooperation this would be a major challenge
for societies in which markets play a central role. Limiting the detrimental effects of market
participation would then require some kind of collective action. However, there are also reasons
to believe that market participation is innocuous or has even beneficial effects on the efficiency
of non-market interactions.
In the field non-market interactions are affected by a multitude of factors which makes it
virtually impossible to tease out the effect of market experience on the basis of field data. The use
of laboratory experiments makes it possible to study spillover effects of market participation with
a high degree of control under ceteris paribus conditions. Specifically, we are able to exogenously
assign participants to the two sides of the market; without laboratory control naturally more
cooperative people might be over-represented on one or the other side. Similarly, we are able
1
Sandel (2012, 2013) discusses how certain exchanges that can be seen as efficiency enhancing are objectionable
on moral grounds and may also crowd out non-market norms. For a survey of different views of market society,
see Fourcade and Healy (2007).
3
to control the composition of the groups in the subsequent social dilemma and, hence, study
behavior for all possible matchings between participants with different market experiences.2
The comparison between these distinct matchings in the social dilemma are the focus of our
work. They will correspond to two distinctions, which are inspired by the work of Henrich et al.
(2001) and by the remarks of Bowles (1998) and capture important variations in market experience. The first is whether participants have previously interacted in the same market or in
different markets, situations to which we will refer to as ‘market-partners and ‘market-strangers,
respectively. The findings of Henrich et al. (2001) suggest that pro-social behavior is enhanced
by market integration with strangers. We explore whether the efficiency of cooperation indeed
depends on whether it takes place between people who interact in the same market or on different markets. This distinction is meant to capture the distinction between more local market,
where the same people compete for scarce resources, and more global market interaction, as for
example in on-line markets.
Second, we also study behavior in the social dilemma for different matchings of people, based
on whether they have been on the long side of the market or on the short side. The distinction
between being on one or the other side of the market corresponds quite directly to the one
proposed by Bowles. It will allow us to explore whether matchings of those who have been
lucky and assigned to the short side of the market, to which we will refer as ‘market-winners,
later behave differently from matchings of those who were unlucky and assigned to the long side,
referred to as market-losers.
In addition, we also take self-assessed measures of experienced well-being and incentivized
measures of social value orientation. These measurements will allow us to study changes in the
subjective well-being and social relations of those on the two sides of the market.
In brief our main results are as follows. For market-partners competitive market experience
has adverse effects on efficient cooperation on both market-winner and market-loser matchings.
In market-strangers matchings of market-winners manage to cooperate more efficiently. These
results indicate that it is not market experience per se that lowers the ability to cooperate. Having
competed for scarce resources on the same side of the market makes it difficult to overcome the
social dilemma, virtually irrespective of having been a market-loser or market-winner. Moreover,
positive market experience can foster cooperation, but only for those who did not have to compete
with each other before. We also show that differences in cooperation cannot be explained by
2
Another advantage of lab experiments is that the possibility of replication allows for a systematic study of the
relevant issues. See Falk and Heckman (2009) for a recent methodological discussion of laboratory experiments.
4
ex-ante income differences and also find evidence that market experience affects subjective wellbeing and social value orientation.
2
Related Experimental Literature
There are a number of related experimental papers, although none of them deals with how
interaction in competitive markets affects subsequent efficiency in cooperation. Brandts et al.
(2009) study the effects of competitive rivalry on the disposition towards others and on subjective
well-being. They focus on pure rivalry and chose a design of a finitely repeated interaction
between fixed triads of players, where one of the three players can in each round choose with
whom of the other two players to play a prisoners dilemma, leaving the third player without
interaction. After the interaction rounds measures of subjects disposition towards others and
experienced well-being were taken. The results show that rivalry has indeed a different impact
on individuals depending on which side of the rivalry they were on. It has a negative impact
on the experienced well-being of those on the long side of the interaction and a positive impact
for the powerful player in the rivalry leading to a larger inequality in experienced well-being.
Interacting under rivalry also affects negatively the disposition towards others.
Bauernschuster et al. (2013) study how competition between two investors interacts with
trust and trustworthiness in simple one-shot trust games. They find that competition among
trustors does not significantly increase sent amounts. However, trustees react to competition
between trustors by lowering return ratios. Similarly, Huck et al. (2012) study a repeated binary
trust game that resembles a market for an experience good with a fixed price where the buyer
can choose whether to trust or not and the seller can only choose quality. Without competition,
buyers are in each period randomly assigned to sellers. With competition, buyers choose in
each period the seller from whom they want to buy. The results show that the introduction of
competition is highly effective, with market efficiency rising from 30 to over 80 percent.
Herz and Taubinsky (2013) study how experience with competition shapes fairness standards.
In their experiment participants first take part in ultimatum games with either proposer or
responder competition and then play the standard ultimatum game. They find that responders
minimum acceptable offers are higher for responders that started in the game with proposer
competition than for those who started in the game with responder competition.
Carpenter and Seki (2006) report on a field experiment conducted with three groups of workers from a fishing community in Japan, where the different groups were exposed to different
5
amounts of competition on-the-job. The results show that these differences explain differences
in cooperation in an experimental setting. Specifically, fishermen and traders, who interact in
more competitive environments are significantly less cooperative than the staff who face little
competition on the job.
Falk and Szech (2013) study behavior in a context in which market exchange can produce
a negative externality – in their case the death of mice. They find that repeated market interaction typically yields less socially responsible behavior than one-shot non-market behavior.
Bartling et al. (2015) study whether markets affect the willingness to incur costs in order to
mitigate a negative externality for third parties. Their study “reveals a persistent preference
among many consumers and firms for avoiding negative social impact in the market.” (p. 219)
However, a direct comparison between behavior in the market context and an individual choice
context reveals that in the former people show less social concern.
Two studies compare the effects of interacting under tournament and under piece-rate incentives of subjects on Amazon Mechanical Turk on subsequent behavior. Buser and Dreber (2013)
find that individuals are significantly less cooperative under tournament incentives than under
piece-rate. In contrast, Chen (2011) finds that “competitively structured work experiences increased non-utilitarian value choices, non-utilitarian commitment towards out-group members”
as well as donations to charity.
Our focus is different from those of previous research. We are interested in exploring the
efficiency effects of the societal experience of being in a market situation where virtually all
market power is on one side of the market whereas the other side has to scramble to obtain the
crumbs of the exchange surplus, as captured in the notion of the reserve army of labor introduced
by Engels (1845).3 In addition, we explore if and how these different market power experiences
interact with whether one has to solve the social dilemma with those one also had to compete
with on the market or with strangers from another market. We believe that the effects of such
situations are of socio-economic importance and need to be studied in depth. We are not aware
of any other experimental work on the matter. In Section 5 we develop specific hypotheses
regarding these effects.
3
Some observers, see e.g. Standing (2011), consider that in modern globalized economies there now exists a
new reserve army of labor, comprised of temporary and part-time workers, who lack any type of job security.
6
3
Experiment Design
Our design has three main building blocks: (1) measurements of subjective well-being and social
value orientation, (2) a highly competitive continuous double auction market (hereafter, DAM)
and (3) a social dilemma game (hereafter, SDG). Our design consists of three treatments, two
treatments in which the DAM is played before the SDG and a control treatment in which the
SDG is played in isolation. The difference between the two treatments involving the DAM, which
we will call the market treatments, consists in the matching of players in the subsequent SDG,
as explained below. All treatments also involve two measurements of subjective well-being and
of social value orientation. We first present the two market treatments followed by the control
treatment.4
3.1
The Two Market Treatments
Both market treatments consisted of eight parts. Figure 1 shows the sequence of events. At
the very beginning, participants were informed that the experiment would have several parts.
Instructions for the various parts were given separately for each part, except those for parts 3
and 4 which were presented together.
Table 1: Sequence of events in market treatments
1.
2.
3.
4.
5.
6.
7.
8.
Self-assessment of subjective well-being
Measurement of social value orientation
Double auction market (18 rounds)
Social dilemma game (6 rounds)
Self-assessment of subjective well-being
Measurement of social value orientation
Surprise restart social dilemma game (12 rounds)
Post-experiment questionnaire
(SWB 1)
(SVO 1)
(DAM)
(SDG)
(SWB 2)
(SVO 2)
(sSDG)
In part 1 (SWB 1) all participants had to answer a self-assessment question to measure their
initial subjective well-being and in part 2 (SVO 1) they had to make money allocation decisions
to measure their social value orientation. In part 3 (DAM) they interacted in 18 rounds of the
DAM and in part 4 (SDG) in six rounds of the SDG. In parts 5 and 6 (SWB 2 and SVO 2,
respectively) participants had again to self-assess their subjective well-being and make money
allocation decisions to measure post interaction social value orientation. Part 7 (sSDG) consisted
of a ‘surprise’ restart of the SDG, lasting for 12 rounds.
4
The experiment instructions can be found in Appendix A.4.
7
Parts 1 and 2: SWB 1 and SVO 1.
In SWB 1 we recorded participants’ response to
the subjective well-being question shown in Figure 1. These initial measurement provides the
baseline to which the second measurement will be compared. Subjects were asked to mark the
number related to the expression of the manikin that best corresponded to how they felt at
that moment.5 In the figure, “1” corresponds to the highest level and “9” to the lowest level of
subjective well-being.
Note: ”1” indicates highest level, ..., ”9” indicates lowest level of subjective well-being
Figure 1: Subjective well-being self-assessment
In SVO 1 we recorded participants’ social value orientation using the so-called circle-test.
The circle-test is a modified and incentivized version of the ring-test (Liebrand, 1984) and was
successfully applied by, among others, Sonnemans et al. (2006) and Brandts et al. (2009). It is
a task which allows for a quantification of individuals’ social value orientation by determining
the readiness of individuals to help or hurt others at some cost to themselves. Figure 2 shows
an example of a circle-test as used in the experiment.
In the circle-test a person’s social value orientation is measured by a single decision which
consists of the selection of a point on a circle. Each point on the circle represents an allocation
S of Experimental Currency Units (ECUs) to the person who makes the choice (Self) and an
allocation O of ECUs to another person (Other). The amounts allocated can be positive or
negative, with S 2 + O 2 = 2002 . Note, that each point on the circle corresponds to a certain
angle, which we will use as the measure of social value orientation. For instance, an angle of 0
degrees corresponds to selfishness as it allocates 200 ECU to oneself and 0 ECU to the other;
an angle of 90 degrees is interpreted as altruistic as it gives 0 to oneself and 200 to the other.
Generally, between 0 and 90 degrees an increasing angle is interpreted as increasing pro-sociality.
A negative angle, which reduces the earnings of the other at some cost to oneself, identifies
competitiveness.6
5
These figures, developed by Lang (1980), are based on Sonnemans (1991).
6
For an extensive discussion of the concept and measurement of social value orientation, see, e.g., Van Lange
(1999) and Murphy et al. (2011).
8
Translation:
With the help of the mouse, choose a point on the circle.
Remember that you can change the decision as many times as
you want.
When you are finished press OK.
NOTE: Remember that you will not interact with the
participant to whom you assign ECUs in any other part of the
experiment and that reciprocity between matched pairs is not
possible.
For myself: 180.11 ECUs
For the other: 86.95 ECUs
Figure 2: Social value orientation circle test
In the experiment the circle appeared on participants’ computer screens. Participants received computerized instructions about how to make the decision and had ample opportunity to
practice.7 Importantly, in the experiment these numbers translate into money earnings at the
exchange rate of 100 ECU to e1. Hence, decisions in the circle-test have pecuniary consequences.
As matched others were random and anonymous, SVO 1 measures the social value orientation
towards generalized others. Subjects were not informed about the decision of ’their others in the
circle-test until the very end of the session.
Parts 3 and 4: DAM and SDG.
Our main interest lies in potential effects of market
experience on efficient cooperation. In order to avoid too much distraction between the two
parts, participants received the instructions for DAM and SDG together. After having read the
instructions and before the start of DAM participants had to answer comprehension questions
about both DAM and SDG.
In the two market treatments participants interacted in the DAM for 18 rounds and in each
round there were the same three sellers and five buyers. Each seller was endowed with two units
of a good which could be sold to the buyers, so that total market supply was six units and total
market demand ten units. That is, buyers were on the long side of the market. The production
7
Each participant, made a social value orientation decision with respect to another anonymously and randomly
chosen participant in the lab. Importantly, the alter-participant does not make a decision towards the egoparticipant but toward yest another randomly chosen participant. This was known to the participants and excludes
(anticipated) reciprocity considerations.
9
cost of each of the units of all three sellers was 10. Each buyer could buy up to two units and the
redemption value of each unit of all five buyers was 100. We chose to give every trader two units
(instead of only one) to create a thicker market with more trades without having to increase the
number of traders. This implied a so-called box design (Holt et al., 1986). We chose that design
because it puts severe competitive pressure on the agents on the long side of the market, in our
case the buyers. This should create distinct market experiences for agents on respectively the
short and the long side of the market.
The earnings from the sale of a unit were equal to the price at which the unit was traded,
while the earnings from the purchase of a unit were equal to 100 minus the price at which the
unit was traded. Not traded units created neither gains nor losses. The price was allowed to
have any integer value between 10 and 95 (inclusive). We chose this upper bound on the trading
price to facilitate trade.8 More formally, in each round the earnings of a buyer in the market
were given by


(100 − px ) + (100 − py ) if the buyer buys one unit at price px





and another unit at price py
u=


(100 − pz )
if the buyer buys one unit at price pz




0
if the buyer does not buy any unit,
and the profit of a seller is given by


(px − 10) + (py − 10) if the seller sells one unit at price px





and another unit at price py
π=
 (pz − 10)

if the seller sells one unit at price pz




0
if the seller does not sell any unit,
where px , py , pz ∈ {10, 11, . . . , 94, 95}. All traders knew their own production cost or redemption
value, but did not know those of the others. Hence, they did not receive any information about
the earnings of the other market participants. Participants were informed about the total number
of buyers and sellers active in the market. We chose this information regime because it has been
shown to minimize fairness considerations and facilitate converge to the competitive equilibrium
(Smith, 1976; Holt et al., 1986). In the competitive equilibrium all six units are traded at price
8
In the experimental literature sometimes trading bonuses are used instead. We did not do that because it
changes the competitive equilibrium prediction. See, e.g., Davis and Holt (1993) and Noussair and Tucker (2013).
10
95. Sellers’ per unit profit is 85 (95 − 10) and buyers’ per unit earnings are 5 (100 − 95).
In the DAM traders had to follow particular trading rules equivalent to those used in previous
double-auction market experiments:
1. Buyers make purchase offers and sellers make sale offers. An offer by a buyer consists in
offering a price at which to buy a unit. An offer by a seller consists in offering a price at
which to sell a unit.
2. Only the highest purchase offer and the lowest sale offer are the so-called pending prices
at which transactions can take place.
3. A transaction takes place automatically if the price of a purchase (sale) offer that is made is
equal or higher (lower) than the price of the pending sale (purchase) offer. The transaction
price is always the pending price (regardless of the offer that leads to the transaction). A
transaction also takes place if a pending purchase (sale) offer is accepted by a seller (buyer).
4. New price offers have to be improvements. That is, a new purchase (sale) offer has to be
higher (lower) than the pending purchase (sale) offer.
5. If a transaction takes place the market clears and any purchase offer or sale offer in the
feasible interval is possible again.
6. The units of the good are traded one by one. That is, traders cannot make offers for or
trade several units at a time.
The DAM was conducted for 18 consecutive rounds with the same fixed group of eight
participants. Participants in a market did not know who they were matched with. A trading
round ended after three minutes or when no trades were possible any more. All participants were
informed about their role in the market, buyer or seller, at the beginning of the 18 rounds of the
DAM and were also told that these roles would stay constant throughout these rounds. During
the DAM buyers and sellers could see the purchase and sale offers and transaction prices but
not the identities behind the offers and transactions. Hence, traders were also not able to track
others’ behavior across market periods. When a trade took place, traders received information
only about their own earnings. At the end of a trading round each trader received information
about his or her total earnings in the round.
After the 18 rounds of the DAM, participants played six rounds of an SDG. The SDG was a
two-person linear public goods game and pairs stayed the same for all six rounds. In each round
each participant was endowed with 50 ECU and had to distribute them between a private and a
public account. We used an MPCR = 0.9 so that for every unit that a player put into the public
11
account both the player in question and the partner’s player obtained 0.9 units.9 Formally, in
each round of the SDG, earnings of a participant i were given by
wi = 50 − gi + 0.9(g1 + g2 ).
In the SDG contribution decisions were made simultaneously. After each participant had made
his/her decision each pair received information about decisions in their pair; that is, own contribution, other’s contribution, own earnings, other’s earnings.
As already mentioned above, the matching in the SDG differed between the two market
treatments: market-partners and market-strangers. In the market-partners (henceforth, MP)
treatment each participant was matched with one of the other seven participant from the same
DAM. Matching was done such that it led to two pairs of two buyers, one pair of sellers and one
pair of one buyer and one seller. Specifically, the instructions specified: “You will be matched
with another buyer (or seller) with whom you have interacted in the market.” Subjects were told
that they would stay matched with the same person during the six rounds of the SDG. In this
way we created one pair of prospective market-winners, two of prospective market-losers and one
mixed pair. For convenience, in the following we will refer to these matchings as buyer-pairs,
seller-pairs, and mixed-pairs.
In the market-strangers (henceforth, MS) treatment each participant in a DAM was matched
with one other participant from another DAM. Here the instructions specified: “You will be
matched with another buyer (or seller) from another market with whom you have not interacted
in the market.” In this case the matchings for the SDG were made using participants from
two different DAMs. The sixteen subjects were matched in a way that led to four buyer-pairs
(prospective market-losers), two seller-pairs (prospective market-winners), and two mixed pairs.
Like in MP the described matchings stayed the same for all rounds of the SDG and participants
were informed about this. Figure 3 provides a graphical representation of the matchings in MP
and MS, respectively.
Parts 5 and 6: SWB 2 and SVO 2.
In SWB 2 we again recorded participants’ response
to the subjective well-being question shown in Figure 1 and in SVO 2 participants again made
9
We used a two-person version because it allowed us to obtain a relatively large number of independent observations at relatively low costs. We chose the MPCR on the basis of two pilot sessions with stand-alone two-person
linear public goods game experiments with the same subject pool as in the reported experiments. In these sessions
we observed that an MPCR = 0.9 lead to efficiency levels of about 50 percent, leaving about the same room for
efficiency improvement and worsening, respectively, in the market experiments.
12
S 11
S 21
S 11
S 21
S 12
S 22
S 12
S 22
S 13
S 23
S 13
S 23
B11
B21
B11
B21
B12
B22
B12
B22
B13
B23
B13
B23
B14
B24
B14
B24
B15
B25
B15
B25
(a) MP
Note:
Sim
(Bjm )
(b) MS
denotes seller (buyer) i (j) in market m.
Figure 3: Matching in the SDG in market-partners (MP) and market-strangers (MS)
decisions in the social value orientation circle-test.10 In SVO 2 each participant made an allocation decision with respect to him/herself and another anonymously and randomly chosen
participant whom s/he did not interact with in any of the previous parts. As in SVO 1, to
avoid (anticipated) reciprocity, the matched participant did not make a decision towards the
deciding participant but towards another not previously matched participant. Again, subjects
were informed about the matching procedure but did not receive information about the decision
of ‘their’ paired others until the very end of the session.
Parts 7: sSDG.
After SVO 2 a ‘surprise’ restart of the SDG was announced and participants
played an additional 12 rounds of the SDG. Each participant was informed that they would be
matched with the same person as in the first six rounds. We introduced the surprise restart to
see if any effects on the efficiency of cooperation due to market experience would be longer lasting
and survive the re-setting commonly observed in public goods games with surprise restarts (see,
e.g., Andreoni, 1988; Croson, 1996).
3.2
Control Treatments
We conducted two control treatments. First, to have a benchmark to which we could compare
contribution behavior in the social dilemma game, with we have run a treatment where partici10
We did not place SWB 2 and SVO 2 directly after the DAM, because we did not want them to influence
behavior in the the SDG which is our main focus.
13
pants played a SDG without having experienced a market before. In that treatment, except for
that there was no DAM, all features of the design were exactly the same as in the above described
market treatments. That is, the sequence of events was the same as depicted in Table 1 (except
for item 3.) and each participant was matched with the same other person both in the first six
and the second 12 rounds of the SDG. We call this treatment Osdg (standing for ‘only’ SDG).
Second, as we will see in the results part, there are – as intended – large earnings differences
between sellers and buyers in the DAM. To control for potential effects of these differences on
contribution behavior in the SDG we conducted another control treatment where participants
received lump-sum payments that corresponded to the average earnings of buyers and sellers in
the two market treatments before they played the SDG. We will call this treatment Osdg-Income.
This treatment was the same as the Osdg save that in Osdg there were no lump-sum payments
before the SDG, and it was the same as the market treatments in terms of (average) pre-SDG
earnings except that participants did not earn these payments through market interactions. The
instructions for the SDG in the control treatment were kept as close as possible to those in DAM.
In particular, regarding the lump-sum payments and the matching with another participant in
the SDG the instructions said: “You have been assigned initial earnings of X ECUs.11 The other
group member is also assigned some initial earnings. The assignments to you and the other group
member are not necessarily the same. You and the other group member will earn this amount
independently of what occurs during the experiment.” We deliberately used a vague phrasing
regarding the earnings of the other group member because in DAM participants also only knew
their own market earnings for sure, whereas they did not have any information about the other
group member’s market earnings. We describe the exact lump-sum earnings and corresponding
matchings in Osdg-Income when we discuss the results (see Section 6.2.1).
4
Experiment Procedures
In total 512 subjects participated in our experiment. We ran three sessions with treatment
Osdg, three with treatment Osdg-Income, three with the market-partners treatment (MP) and
four with the market-strangers treatment (MS). We have data from 64 subjects in Osdg, 144
in Osdg-Income, 112 in market-partners in 14 separate markets, and 192 in market-strangers
in 24 separate markets. Hence, for Osdg we have 32 statistically independent pairs and for
Osdg-Income 72, which are distributed over six different conditions with 12 independent pairs
11
The amount shown depended on the condition the participant was assigned to.
14
per condition (see Section 6.2.1 for details). For market-partners we have 56 pairs in the SDG
(28 buyer-pairs, 14 seller-pairs, 14 mixed-pairs) organized in 14 independent matching groups
(markets) and for market-strangers we have 96 pairs in the SDG (48 buyer-pairs, 24 seller-pairs,
24 mixed-pairs) organized in 12 independent matching groups (twins of markets).
In the two market treatments, each participant’s role (buyer or seller) was fixed for the
duration of the session. Instructions and a careful explanation were read aloud at the start of
each session.Participants could ask questions by raising a hand. All questions were answered in
private.
The experiments were conducted at the LINEEX lab at the University of Valencia using the
z-tree program of Fischbacher (2007). Each session involved only one of the treatments. No one
could participate in more than one session. Total performance-based earnings were counted in
ECU and consisted of the accumulated earnings across all parts. Each 100 ECU were worth e1.
Participants did not receive a show-up fee. At the end of a session participants were privately paid
out their earnings in cash. Average earnings were e18.50 for the Osdg, e33.00 for Osdg-Income,
and e29.50 for the market treatments. Osdg sessions took about 90 minutes, Osdg-Income about
XX minutes, and market treatments sessions took about 120 minutes.
5
Research Hypotheses
Our main focus is on finding out understanding to what extent participation in a competitive
market has repercussions for the efficiency achieved in subsequent social dilemma games. In
addition, we are also interested in studying the mechanisms that may lead to an interaction
between the two spheres using our measurements of subjective well-being and of social value
orientation.
We formulate our hypotheses inspired by some of the concepts developed by
Bowles and Polania-Reyes (2012). The focus of their study is the analysis of whether economic
incentives and social preferences are substitutes or complements. Two key elements of their
conceptual framework are the notion of separability between economics incentives and social
preferences and the distinction between state-dependent and endogenous preferences.12
Applied to our context separability means that experiences in different spheres of social
12
Bowles and Polania-Reyes (2012) make the distinction between state-dependent and endogenous preferences,
where the latter term is used in relation to processes with effects that persist in the long-run typically as the result
of a process of cultural transmission. In the context of our study, the effects we focus on can be better captured
in terms of state-dependence.
15
interaction do not interact with each other and, more specifically, that the ability to cooperate in a
social dilemma game is independent of preceding market experience. The notion of separability is
implied in the view proposed by Smith (1998) that people are able to decouple behavior in smallgroup exchange from that in anonymous markets. In our experimental environment separability
of market experience and cooperation in a non-market environment is consistent with the absence
of effects of market experience on the efficiency of cooperation as well as subjective well-being
and social value orientation. Formally our null hypothesis is:
Hypothesis 1. Market experience has neither an effect on cooperation in the SDG nor subjective
well-being or social value orientation.
Our alternative hypotheses are formulated in terms of state-dependence of behavior in the
SDG. Behavior has been found to depend on the circumstances surrounding the decision situation
and Bowles and Polania-Reyes (2012) propose the notion of state-dependence to conceptualize
some of this behavior. “State-dependence arises because actions are motivated by a heterogeneous repertoire of preferences from spiteful to payoff-maximizing to generous, for example the
salience of which depends on the nature of the decision situation.” (p. 373).13
The general alternative to our null hypothesis is that behavior will be state-dependent in the
sense that distinct market experiences will trigger different attitudes towards subsequent cooperation. However, state-dependence may affect behavior in different ways along two dimensions.
First, market interaction as such may affect behavior in the subsequent SDG. The effect could
be to increase cooperation, in accordance with the idea of doux commerce of Montesquieu (1748)
or it could be negative, in line with the social criticism of Engels (1845).
Second, state-dependence could lead to variations in reactions depending on the type of
market experience. In particular, having being on the short or the long side of the market
and being paired with somebody fro the own or the other side may lead to different reactions.
Similarly, cooperation may depend on whether one is paired with somebody one has previously
competed for scarce resources with in the market or not.
A final issue pertaining to state-dependence are the mechanisms that could be behind it. In
Brandts et al. (2009) we found that being on the long side of a triadic exchange relation leads to
13
An example of how state-dependence could be incorporated into a formal model of social preferences is the
general model of Charness and Rabin (2002). This two-person model has a more standard part with own and
others payoff and also incorporates a particular parameter that is said to be set to 1 when the decision-maker
thinks that the counter-part is misbehaving while it is set to 0 when the counter-part is not misbehaving. The
state is whether the counter-part is misbehaving or not and this gives rise to a repertoire of two different social
preferences.
16
lower subjective well-being and less social value orientation, measured after the exchange relation
had finished. The issue here is whether possible effects of market experience on cooperation can
be related to effects on subjective well-being and social value orientation.
We formulate alternative hypotheses in terms of the effects on cooperation. Our analysis of
subjective well-being and social value orientation will be part of a more detailed discussion of
these alternatives.
Hypothesis 2. Market participation as such will affect subsequent cooperation.
Hypothesis 2 is two-sided, reflecting the existence of conflicting ideas on the effects of market
participation, as discussed above. In evaluating it we will simply compare SDG behavior with
and without earlier market interaction.
Importantly, our design allows us to make a number of more specific comparisons of interest by looking at behavior across several cases. First, we can compare market-strangers with
market-partners. As mentioned in the introduction, these two treatments are meant to capture
the distinction between more local market interaction and more distant market interaction. One
may expect the different kinds of relations in the market to affect participants’ attitudes towards the subsequent interaction. In market-partners the experience of having competed with
each other for scarce resources may on the one hand inject some sense of social closeness and
thus increase cooperation, but on the other hand it may also induce a competitive state that
could be detrimental to efficient cooperation. Market-strangers may create an atmosphere of
more anonymity and disconnectedness. Whether this will have a positive or negative effect on
cooperation is an open question.
Second, we can hold the market-strangers vs. market-partners distinction constant and compare behavior in different matchings of participants who have been on the same or on opposite
sides of the market. In this case, the issue is whether pairs of participants who both have been
assigned to the long side of the market, market-losers, to cooperate better or worse than pairs
with other market experiences and pairs with no market experience. It is not easy to gauge
what the effects could be. Given that market-losers earn little income, one many expect pairs
of them to cooperate more effectively out of a desire of compensating for low earnings in the
market. Also, a sense of solidarity between them may lead to better cooperation. However, bad
market experience could also have a negative effect on market-losers by depleting their trust in
others or triggering spiteful behavior. Market-winners have a good experience, at least in terms
of earnings, which may on the one hand make it easier to cooperate but on the other hand also
17
here a competitive state of participants may affect cooperation negatively.
Under separability none of these comparisons should exhibit significant differences. However,
as just described there are reasons suggesting opposing forces under the different circumstances
and it is an empirical question, which one – if any – will prevail. We formulate this in our second
general two-sided alternative hypothesis.
Hypothesis 3. The type of experienced market participation will affect subsequent cooperation
in different ways.
This hypothesis can be split into a number of specific questions. How does market participation affect the efficiency of post-market cooperation of pairs of individuals:
(A) who had interacted in the same market (market-partners) compared to pairs of individuals
who had interacted in distinct markets (market-strangers) and to those without market
experience?14
(B) who had both been on the favorable side of markets (market-winners) compared to pairs
of individuals who had both been on the unfavorable side of markets (market-loser) and
compared to pairs of individuals who had been on the opposite side of markets, all three
compared between market-partners and market-strangers?
(C) who had both been on the favorable side of markets compared to pairs of individuals who
had both been on the unfavorable side of markets and compared to pairs of individuals who
had been on the opposite side of markets, all three compared within market-partners and
market-strangers?
Brandts et al. (2009) found that experiencing competitive rivalry in fixed triads had a negative impact on the experienced well-being of those on the long side of the interaction and a
positive impact on those on the short side. Here we study whether this result extends to the
case of market interaction in a proper competitive market, and whether it is affected by the
distinction between market-partners and market-strangers.
(A) Is experienced well-being, as measured by a subjective self-evaluation, affected by market
participation and by the side of the market a person is on?
14
In this comparison we average the different types of matchings. We leave out the other possible comparison
of averages, where we would average for each type of matching across market-partners and market-strangers.
18
(B) Is social value orientation affected by market participation and by the side of the market a
person is on?
For social value orientation Brandts et al. (2009) found that the powerful player in the triad
did not change his disposition towards others (i.e., social value orientation) towards the player
that had been chosen more frequently for interaction but surprisingly lowered it with respect to
the player less frequently chosen. Both powerless players lowered their social value orientation
towards others.
In our set-up, if market participation has a negative effect per se, then social value orientation
should be expected to lower the disposition of all market participants. In contrast, if social value
orientation responds to income comparisons, as e.g. in Fehr and Schmidt (1999), then social
value orientation of those on the short side of the market should be expected to increase (or
at least not decrease) while that of those on the long side of the market should be expected to
decrease (to compensate for the relative income disadvantage). In addition, we will be able to
investigate the effect of the market-partners vs market-strangers distinction.
6
Results
In this section, we first briefly report on market behavior to see if our markets indeed converge to
asymmetric outcomes as intended. Then, we zoom in to our main research question and discuss
if and how different market experiences affect behavior in the subsequent social dilemma games.
Thereafter, we discuss the effect of different experiences in the markets and social dilemma games
on subjective well-being and social value orientation.
6.1
Market Behavior
Figure 4 shows the average trading price over the 18 trading rounds in the two market treatments.
As expected, prices in both treatments converge to the highest possible price of 95. Of the total
of 4104 possible trades only 7 were not realized and overall efficiency was with 99.8% virtually
optimal. This market outcome leads to very unequal incomes. Using individual data the averages
and standard deviations of earnings are 2672 (stdev: 277) for sellers and 340 (stdev: 176) for
buyers in the market-partners treatment and 2656 (324) for sellers and 346 (222) for buyers in the
market-strangers treatment.15 As expected, neither buyer nor seller earnings significantly differ
15
Note that if there are any pre-existing social preferences they apparently have little effect on the outcome of
the market interaction, due to the competitiveness of the institution (see Bolton and Ockenfels, 2000).
19
100
average price in treatment
90
80
70
60
50
40
30
20
10
0
1
2
3
4
5
6
7 8 9 10 11 12 13 14 15 16 17 18
trading period in market
Market−partners
Equilibrium price
Market−strangers
Figure 4: Average trading price dynamics in both market treatments
between market-partners and market-strangers treatments (buyer earnings: p = 0.6434, sellerearnings: p = 0.5371; MW-tests, 2-sided). We can conclude that our manipulation worked as
intended. With the markets we established the pre-condition of very different market experiences
for participants on the long and on the short side of the market, without differences between the
two market environments.
6.2
Behavior in the Social Dilemma Game
Figure 5 shows mean contributions aggregated for each of the 18 rounds in the SDG of the
two market treatments and the Osdg treatment. In rounds 1-6 there is a clear ranking of the
treatments visble. Contributions in the market-strangers treatment are above those in the control
treatment without market interaction (except for the last round), which in turn are above the
ones in the market-partners treatment. In rounds 7-18 contributions in the market-strangers
and Osdg treatments are similar, while in the market-partners treatment they are clearly below
those of the other two treatments.16
We organize our statistical analysis, treating our three main research questions one by one.
Using regressions, Table 2 reports tests on differences across treatments in contributions, aggre16
In all treatments we observe the well-known end-of-game effect, both in round 6 and in round 18. Observe
also that the data for all three treatments exhibit a restart effect, first reported in Andreoni (1988); that is, in
round 7 contribution levels jump up to levels similar to those of rounds 1-5.
20
50
45
mean contribution
40
35
30
25
20
15
10
5
0
1
2
3
4
OSDG
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
Market−partners
Market−strangers
Figure 5: Average contributions over rounds in the two market treatments and the Osdg treatment
gated over trader matchings and separately for all rounds, for rounds 1-6, and for rounds 7-18.17
In all reported specifications contributions are regressed on dummy variables identifying the
market-partners and market-strangers treatment, respectively. Treatment Osdg is the omitted
category.18 Specification (1) reports the results for all 18 periods and shows that contributions
in market-partners are 6.92 points lower than in Osdg (p = 0.036). Corroborating the visual
impression from Figure 5, an F -test comparing contributions in market-partners and marketstrangers shows that contributions are highly significantly lower in the former than in the latter
(diff.: 8.298, p < 0.001). There is no statistically significant difference between contributions in
market-strangers and Osdg (p = 0.622). Specifications (2) and (3) in Table 2 show a very similar
picture for rounds 1-6 and 7-18 separately. In market-partners contributions are significantly
lower than in the other treatments. Interestingly, as already suggested by Figure 5, comparing
contributions in rounds 1-6 and rounds 7-18 indicate that the negative effect in market-partners
does not vanish, but is rather reinforced over time.
17
In the main text, we report statistical tests using OLS regression analysis based on individual contributions
with standard errors corrected for dependence of observations within pairs in Osdg, within individual markets in
market-partners, and within pairs of interlinked markets in market-strangers. The results are largely consistent
with those of non-parametric tests, which are reported in Appendix A.1.
18
For all reported regressions here and later in the paper we have also run specifications adding period dummies
to control for time trends and Tobit instead of OLS to correct for censoring of contributions by 0 and 50. The
results are virtually the same and reported in Appendix A.2.
21
Table 2: Contributions in the social dilemma game in the three treatments (aggregate across
trader matchings)
Specification
Market-partners
Market-strangers
Constant (Osdg)
R2
N
F(1,57)
p-value
dep. var.: contribution to public good
(1)
(2)
(3)
all periods periods 1-6
periods 7-18
-6.919∗∗
-5.425∗
-7.665∗∗
(3.228)
(3.249)
(3.424)
1.379
2.968
0.585
(2.779)
(2.734)
(2.990)
29.375∗∗∗
28.451∗∗∗
29.837∗∗∗
(2.640)
(2.589)
(2.815)
0.037
0.039
0.036
6624
2208
4416
MPM vs MSM
16.37
15.21
14.14
0.0002
0.0003
0.0004
Note: ∗∗∗ , (∗∗ ), [∗ ] indicated significance on 1, 5, and 10 percent level,
respectively. Standard errors corrected for dependence of observations for
58 clusters.
Hence, overall market participation can be harmful for cooperation. However, it is not market
experience per se what affects cooperation, but having to cooperate with people with whom one
has previously interacted in the market. Not having competed in the same market - something
that can be seen as a more anonymous environment – leads to higher cooperation than having
had to compete with each other. The differences survive the restart after the first six rounds and
even become somewhat stronger. We summarize this answer to our research question pertaining
to overall cooperation levels in the following result.
Result 4. Market participation harms efficient cooperation, but only when agents had competed
in the same market before.
We next move to research question, whether the aggregate differences just discussed are
similar for the different trader matchings or whether they are driven by specific matchings.
Recall that in buyer-buyer matchings the two interacting participants both have had difficulties
securing trades and have obtained low earnings. In seller-seller matchings participants have
technically also competed for trades but from a very comfortable position and have obtained
high earnings. Finally, buyer-seller matchings bring together very different market experiences
and earnings levels.
Figure 6 shows mean contribution levels for the three distinct kinds of matchings in the two
22
market treatments as well as in the stand-alone Osdg. Figure 6a depicts contributions in buyerbuyer pairs and clearly shows that buyers do worst in market-partners in all rounds (expect
the very last one), whereas contributions are similar in market-strangers and Osdg. Regarding
market-partners, a similar picture emerges for seller-seller pairs (Figure 6b). Their contributions
are clearly below those of sellers in market-strangers and of agents in Osdg. Interestingly, sellerseller pairs seem to contribute more than agents in Osdg after having experienced the marketstrangers environment. Finally, for seller-buyer pairs, Figure 6c shows little differences across
50
50
45
45
40
40
35
35
mean contribution
mean contribution
treatments.
30
25
20
15
30
25
20
15
10
10
5
5
0
0
1
2
3
4
OSDG
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
Market−partners
1
2
Market−strangers
3
4
5
6
7
OSDG
(a) Buyer-buyer & Osdg
8
9 10 11 12 13 14 15 16 17 18
period
Market−partners
Market−strangers
(b) Seller-seller & Osdg
50
45
mean contribution
40
35
30
25
20
15
10
5
0
1
2
3
4
OSDG
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
Market−partners
Market−strangers
(c) Seller-buyer & Osdg
Figure 6: Average contributions over rounds in the three treatments by different trader matchings
Table 3 reports regression analysis across all periods, equivalent to Specification (1) of Table 2,
but now separately for the three different trader matchings in the market treatments. Again,
Osdg is the omitted category. The results shown in the table corroborate the visual impressions
gained from Figure 6. For buyer-buyer matchings, contributions in market-partners are 9.68
23
Table 3: Contributions in the social dilemma game in the three treatments by trader matchings
(all periods)
Specification
Market-partners
Market-strangers
Constant (Osdg)
R2
N
F(1,57)
p-value
dep. var.: contribution to public good
(1)
(2)
(3)
buyer-buyer seller-seller
seller-buyer
-9.678∗∗
-8.319∗
-0.000
(3.768)
(4.254)
(4.749)
-0.970
5.899∗
1.557
(3.126)
(3.288)
(3.433)
29.375∗∗∗
29.375∗∗∗
29.375∗∗∗
(2.640)
(2.640)
(2.640)
0.045
0.073
0.001
3888
2520
2520
MPM vs MSM
7.56
13.51
0.12
0.0080
0.0005
0.7316
Note: ∗∗∗ , (∗∗ ), [∗ ] indicated significance on 1, 5, and 10 percent level,
respectively. Standard errors corrected for dependence of observations for
58 clusters.
points lower than in Osdg and 8.71 points lower than in market-strangers. Both differences
are statistically significant (p = 0.013 and p = 0.008, respectively). For seller-seller pairs we
again find that contributions in market-partners are significantly lower than in market strangers
(diff.: 14.22, p < 0.001). This difference is partly due to the fact that seller-seller pairs in the
market-strangers treatment contribute even more than agents in Osdg (diff.: 5.90, p = 0.078).
Finally, for the buyer-seller matchings there are no statistically significant differences across
treatments (p ≥ 0.652).
By and large, the differences identified when taking all periods into account, are also found
when looking at rounds 1-6 and rounds 7-18 separately (see Table 4). For buyer-buyer pairs
contributions are lower in market-partners than in both market-strangers and Osdg already in
periods 1-6 (diff.: 7.51, p < 0.010 and diff.: 6.72, p = 0.056, respectively). These differences do
not vanish but rather increase in the later periods 7-18 (diff.: 9.31, p = 0.012 and diff.: 11.16, p =
0.010, respectively). Between market-strangers and Osdg contributions significantly differ neither
in the early nor the later periods. For seller-seller pairs in periods 1-6 contributions are lower in
market-partners than in market strangers (diff.: 13.72) and Osdg (diff.: 6.68), which is highly
significant for in the former (p < 0.001) but fails to be statistically significant in the latter (p =
0.112) comparison. For the later periods 7-18 both differences increase and become statistically
(marginally) significant (diff.: 14.47, p = 0.001 and diff.: 9.14, p = 0.056, respectively). Seller24
seller pairs contribute more in market-strangers than agents in Osdg, especially in the early
periods (periods 1-6: diff.: 7.04, p = 0.038; periods 7-18: diff.: 5.33, p = 0.131). For seller-buyer
pairs contributions significantly differ neither in periods 1-6 nor in periods 7-18 (p ≥ 0.289).
Table 4: Contributions in the social dilemma game in the three treatments by trader matchings
(periods 1-6 & periods 7-18)
Specification
Market-partners
Market-strangers
Constant (Osdg)
R2
N
F(1,57)
p-value
Specification
Market-partners
Market-strangers
Constant (Osdg)
R2
N
F(1,57)
p-value
Periods 1-6
dep. var.: contribution to public good
(1)
(2)
(3)
buyer-buyer seller-seller
seller-buyer
-6.721∗
-6.682
-1.576
(3.447)
(4.140)
(4.794)
0.786
7.036∗∗
3.265
(3.062)
(3.318)
(3.308)
28.451∗∗∗
28.451∗∗∗
28.451∗∗∗
(2.590)
(2.591)
(2.591)
0.031
0.073
0.010
1296
840
840
MPM vs MSM
7.18
12.78
1.14
0.0096
0.0007
0.2895
Periods 7-18
dep. var.: contribution to public good
(1)
(2)
(3)
buyer-buyer seller-seller
seller-buyer
-11.156∗∗∗
-9.138∗
0.788
(4.173)
(4.681)
(5.069)
-1.848
5.331
0.703
(3.380)
(3.479)
(3.782)
29.837∗∗∗
29.837∗∗∗
29.837∗∗∗
(2.815)
(2.816)
(2.816)
0.054
0.073
0.000
2592
1680
1680
MPM vs MSM
6.67
11.53
0.00
0.0124
0.0013
0.9862
Note: ∗∗∗ , (∗∗ ), [∗ ] indicated significance on 1, 5, and 10 percent level,
respectively. Standard errors corrected for dependence of observations for
58 clusters.
We briefly summarize in our next result.
25
Result 5. The observed adverse effects of market experience on efficient cooperation in marketpartners can be attributed both to buyer-buyer and seller-seller matchings. For seller-seller pairs
market experience in market-strangers tend to positively effect contribution behavior.
This suggests that having experienced the severity of competition in the experimental market
together negatively affects agents’ ability to successfully cooperate. Importantly, this effect does
not vanish over time and is present both for partnerships of market losers and for partnerships of
market winners. This implies that this phenomenon can not purely be explained by the different
income levels attained in the market. Interestingly, market winners in the more socially distant
market-strangers treatment are able to increase contributions relative to the control treatment.
This suggests that market experience may be beneficial for cooperation provided the market
experience is positive and not with the same people one has to solve the social dilemma with.
This seems consistent with the indirect evidence provided in Camerer and Fehr (2004).
In our analysis up to now, we have established the effects of different market experiences
on the efficiency of cooperation. In doing so we also saw that in market-partners the observed
adverse effect is due to worse cooperation of both market-losers (buyer-buyer pairs) and marketwinners (seller-seller pairs). Moreover, in market-strangers it appeared that seller-seller pairs’
contribution behavior is positively affected by market experience. This suggests that in marketpartners there is little to no difference in cooperation between market-losers and market-winners
while in market-strangers seller-seller pairs contribute more than other pairings. In the following
we test for these differences in contributions between different trader pairings, separately for
both market treatments. Table 12 reports the results (aggregated over all periods) of F -tests
based on regression analysis using dummy variables for respectively buyer-buyer, seller-seller,
and seller-buyer pairs, with treatment Osdg being the omitted category.19
For the market-partners treatment we indeed see that buyer-buyer and seller-seller pairs do
not contribute differently (p = 0.762)) and that both contribute less than mixed trader pairs.
The latter difference is statistically significant for buyer-buyer pairs (p = 0.0433) but not for
seller-seller pairs (p = 0.1150). For the market-strangers treatment, also as expected, we find
that pairs of market winners (seller-seller) contribute more than buyer-buyer and mixed trader
pairs. The difference is statistically significant when comparing to market-loser pairs (p = 0.0207)
but not when comparing to mixed trader pairs (p = 0.1365).20 We summarize in our next result.
19
For convenience we report here only the comparison results. The whole regression table can be found in
Appendix A.2, Table 12.
20
The results are similar when looking at periods 1-6 and 7-18 separately and reported in Appendix A.2, Table 13.
26
Table 5: Contributions in the social dilemma game: comparing trader matchings within market
treatments (all periods)
diff. contributions
F
p-value
diff. contributions
F
p-value
diff. contributions
F
p-value
Market-partners Market-strangers
buyer-buyer - seller-seller
-1.359
-6.869
0.09
5.77
0.7624
0.0207
buyer-buyer - seller-buyer
-9.678
-2.527
4.32
0.62
0.0433
0.4356
seller-seller - seller-buyer
-8.319
4.342
2.58
2.30
0.1150
0.1365
Note: F1,45 (F1,43 ) for market-partners (market-strangers).
Result 6. When there was a joint market experience (market-partners) both market-loser pairs
and market-winner pairs contribute similarly and both (tend to) contribute less than pairs consisting of a winner and a loser. In markets without a joint experience (market-strangers) pairs of
market-winners (tend to) contribute more than pairs consisting of market-loser or mixed pairs.
6.2.1
Controlling for the Effect of Income Differences
An inherent property of markets with a short and a long market side is that market opportunities and thus earnings are asymmetric. We deliberately implemented such markets with
highly unequal opportunities for market participation because it reflects important markets in
the field, as argued in the introduction. Buyers and sellers in our markets therefore ended up
with very different earnings after the market phase and one may argue that the differences we
find in the social dilemma game are not due to different market experiences but solely a result of
different earnings. In other words, it may be that differences in contributions to the public good
are merely (or mostly) a reflection of income effects. The result that in the market-strangers
treatment seller-seller pairs, which had high market incomes, cooperate most efficiently indeed
points to such a possibility. On the other hand, however, results in market-partners are not
consistent with this interpretation as there both seller-seller, pairs with high market earnings,
and buyer-buyer pairs, with low market earnings, cooperate rather inefficiently.
To directly test for income effects we have run the control treatment Osdg-Income where
27
participants did not participate in markets but started with the same average income differences
as in our market-partner and market-stranger treatments (see also Section 6.2.1). Recall from
Section 6.1 that in market-partners on average sellers earned 2672 ECU and buyers 340 ECU and
in market-strangers the corresponding earnings were 2656 ECU and 346 ECU. In Osdg-Income
we assigned participants to the following lump-sum earnings pairs: 340-340, 2672-2672, 2672-340
to mimic matchings in market-partners and 346-346, 2656-2656, 2656-346 to mimic matchings in
market-strangers. In the following we explore if contributions of these pairs are in line with the
SDG contributions of different matchings observed in the market treatments. Specifically, we are
interested if the lump-sum earnings matchings 340-340 and 2672-2672, respectively, can explain
the relatively low contributions of respectively buyer-buyer and seller-seller pairs in marketpartners. Similarly, we are particularly interested in whether the relatively high contributions
of seller-seller pairs in market-strangers could be explained by the high earnings of 2656-2656
matchings in Osdg-Income.
To get a first impression Table 6 reports descriptive statistics of contributions over all periods
for the different matchings in market-partners, market-strangers, the corresponding matches in
Osdg-Income, and as a benchmark Osdg. The statistics reported in the table already suggest
that the differences in contributions are likely not due to different earnings participants hold
before they interact in the social dilemma game. First, when looking at contributions across
different matchings within treatment Osdg-Income we see little differences between high, low
and mixed earnings pairs. They range from 29.03 high-earnings pairs 2656-2656 to 33.31 in
low-earnings pairs 346-346. Second, the low contributions in buyer-buyer and seller-seller pairs
in market-partners (19.70 and 21.06, respectively) are not accompanied by low contributions in
the corresponding 340-340 and 2672-2672 earnings pairs in Osdg-Income, where contributions
amount to 33.44 and 32.51. Third, the relatively high contributions of seller-seller pairs in market
strangers (35.27) are not matched with higher than average contributions in the corresponding
2656-2656 earnings pairs in Osdg-Income (29.03).
To statistically test for differences in contributions we have run regression analysis similar to
the ones reported below. Specifically, we regressed contributions on dummy variables representing the different matchings in respectively the market-partners and market strangers treatments
and the different corresponding lump-sum earnings matchings in treatment Osdg-Income, with
Osdg as the omitted category. We then used F -tests to examine statistically significant differences. In the following we report the main results. The regression table with all tests can be
found in Appendix A.2, Table 14.
28
Table 6: Average contributions (with standard deviations) in the social dilemma game for all
treatments by matchings (all periods)
Market-partners
buyer-buyer
19.70
(18.34)
seller-seller
21.06
(18.29)
seller-buyer
29.38
(20.12)
Osdg-Income
340-340
33.44
(19.11)
2672-2672
32.51
(18.49)
2672-340
30.38
(20.79)
Market-strangers
buyer-buyer
28.41
(18.45)
seller-seller
35.27
(17.20)
seller-buyer
30.93
(20.06)
Osdg-Income
346-346
33.31
16.67
2656-2656
29.03
(18.41)
2656-346
29.46
(20.37)
Osdg
29.38
(18.70)
Note: Standard deviations in parenthesis. Statistics are based on individual contributions
in each period.
We first test if contributions significantly differ across different income matchings within
Osdg-Income as such differences would be a first indication that pre-SDG earning differences
may affect contributions in the SDG. The descriptive statistics above suggest that this is not the
case and the statistical analysis corroborates this impression. Equality of contributions across
different lump-sum earnings matchings cannot be rejected neither for those corresponding to
the market-partners treatment (p ≥ 0.5917) nor for those corresponding to the market-partners
treatment (p ≥ 0.3581). Second, to explain the lower contributions for buyer-buyer and sellerseller pairings in market-partners by the earnings achieved in the market interactions we should
see similar contribution behavior in the corresponding 346-346 and 2672-2672 matchings. This
is not the case as in both comparisons contributions are significantly lower in market-partners
than in Osdg-Income (p = 0.0026 and p = 0.0305). Finally, we test if the higher seller-seller
contributions in market-strangers are the same as contributions in 2656-2656 matchings. In
this case, although seller-seller contributions are higher than 2656-2656 contributions (35.27 vs
29.03), we cannot reject equality of contributions. On the other hand, however, the contributions
of 346-346 matchings are with 33.31 closer to those of seller-seller pairs. Thus, there is no evidence
that higher pre-SDG earnings are the cause for the higher contributions of seller-seller pairs in
market-strangers. We summarize in or next result.
Result 7. The differences in contributions in the social dilemma game observed in the market
treatments cannot be attribute to the earnings differences generated in the market interactions.
29
6.3
Subjective well-being and social value orientation
We now first discuss if and how market experience affect subjective well-being followed by a
discussion of the effect on social value orientation. Recall that we asked participants to respond
to our subjective well-being question at two points during the experiment: in Part 1, at the very
beginning of the experiment (initial subjective well-being), and in Part 5, that is after the market
interaction (in the market treatments) and first six rounds of the SDG had taken place (final
subjective well-being). Table 7 depicts descriptive statistics of the initial and final subjective
well-being measure as well as the change form the first to second measurement point. As the
initial subjective well-being differs across participants we focus on the change in this measure
(see bottom of Table 7). We see a general decline in subjective well-being in all three treatments
and this decline appears to be stronger in the market treatments than in the control treatment
Osdg.
Table 7: Subjective well-being and social value orientation (descriptive statistics)
Treatment
Market-partners
Market-strangers
Osdg
Initial subjective well-being
Mean St.Dev.
N
3.196
1.749
112
2.911
1.594
192
2.922
1.525
64
Initial
Mean
11.571
17.659
21.876
social value orientation
St.Dev.
N
26.660
112
20.386
192
21.164
64
Treatment
Market-partners
Market-strangers
Osdg
Final
Mean
3.696
3.641
3.156
subjective well-being
St.Dev.
N
2.265
112
2.347
192
1.720
64
Final
Mean
9.124
11.627
17.338
social value orientation
St.Dev.
N
19.312
112
21.165
192
21.512
64
Treatment
Market-partners
Market-strangers
Osdg
Change
Mean
-0.500
-0.729
-0.234
in subjective well-being
St.Dev.
N
2.344
112
2.599
192
1.771
64
Change
Mean
-2.447
-6.031
-4.538
in social value orientation
St.Dev.
N
17.619
112
19.259
192
13.682
64
Note: For subjective well-being a higher value indicates worse well-being;
change in subjective well-being is initial - final. For social value orientation higher degrees (smaller 90 degree) indicate stronger pro-social value
orientation; change in social value orientation is final - initial.
To test if these differences are significant we have conducted regression analysis with the
change in subjective well-being as the dependent variable and as independent variables the
30
treatment participants have been in (dummy variable ‘Market-partners’ and ‘Market-strangers’
with Osdg being the omitted category) and individual earnings in the first six periods of the SDG
(‘Earnings in SDG 1’). Specification (1) in Table 8 shows that there is a general statistically
significant decline in subjective well-being (see Constant) that is significantly stronger for those
who experienced the market-strangers treatment (p = 0.024). For market-partners the coefficient
is also negative but statistically not significant (p = 0.739). Between the two market treatments
the difference in the decrease in subjective well-being is statistically marginally significant at
p = 0.060. Not surprisingly earnings in the SDG have a highly significant positive impact on
subjective well-being (p ≤ 0.001).
Table 8: Regression: change in subjective well-being
Specification
Earnings in SDG 1
Market-partners
Market-strangers
Buyer
Seller
Market-partners*Buyer
Market-partners*Seller
Market-strangers*Buyer
Market-strangers*Seller
Constant (Osdg)
R2
N
F(1,57)
p-value
dep. var.: change in subjective well-being
(1)
(2)
(3)
∗∗∗
∗∗
0.007
0.005
0.005∗∗∗
(0.002)
(0.002)
(0.247)
-0.087
(0.259)
-0.593∗∗
(0.256)
-1.130∗∗∗
(0.285)
0.796∗∗
(0.243)
-0.614∗
(0.333)
0.672∗∗
(0.334)
-1.424∗∗∗
(0.335)
0.859∗∗∗
(0.248)
-3.236∗∗∗
-2.237∗∗
-2.475∗∗∗
(0.664)
(0.714)
(0.660)
0.039
0.155
0.1691
368
368
368
MPM vs MSM Seller vs Buyer
all comp. sig.
3.68
43.96
except
0.0599
0.0000
MSM*Seller vs MPM*Seller
Note: ∗∗∗ , (∗∗ ), [∗ ] indicated significance on 1, 5, and percent level, respectively. Standard errors corrected for dependence of observations for
58 clusters.
31
Table 9: Change in subjective well-being: comparison of market roles
F(1,57)
p-value
F(1,57)
p-value
F(1,57)
p-value
MP*buyer=MP*seller
8.36
0.0054
MP*buyer=MS*buyer
4.21
0.0448
MP*buyer=MS*seller
21.37
0.0000
MS*buyer=MS*seller
44.58
0.0000
MP*seller=MS*seller
0.38
0.5427
MP*seller=MS*buyer
28.31
0.0000
Specification (2) of Table 8 distinguish between buyers and sellers (and for the moment
ignore which market they have been in). The results for this regression show that sellers report a
significant increase in their subject well-being (p = 0.002), while buyers experience a significant
decrease (p ≤ 0.001). The difference between the coefficients is highly significant (p ≤ 0.001).
Also in this specification earnings in the SDG have a positive impact on subjective well-being
(p ≤ 0.007).
Finally, specification (3) shows that the effect on buyers is (marginally) significantly negative both for market-partners (p = 0.066) and for market-strangers (p < 0.001), whereas it is
significantly positive for sellers irrespective of the type of experienced market (market-partners:
p = 0.066, market-strangers: p < 0.001). The differences between buyers and sellers are all
statistically significant (p ≤ 0.005; see Table 9).
We summarize the results pertaining to the effect of market experience on subjective wellbeing as follows:
Result 8. Subjective well-being is generally decreasing over time. In both market environments,
it decreases most strongly for market-losers (buyers) and least strongly for market-winners (sellers).
The panels on the right side of Table 7 report descriptive statistics of our measure of social
value orientation measured before any market and SDG experience (Initial social value orientation) and after market experience and the first sir period of the social dilemma game (Final
social value orientation). The descriptive statistics reported in the table show quite some variation across treatments as well as individuals. Therefore, as before, we focus on change in social
value orientation from the first to the second measurement point (lower right panel in Table 7).
Consistent with the literature (Sonnemans et al., 2006; Brandts et al., 2009), social value ori32
entation exhibits a general decline. Interestingly, relative to the benchmark Osdg the decline
appears stronger for participants in market-strangers and weaker for those in market-partners.
To text for statistical significance we have run regressions similar to those for subjective wellbeing. Table 10 reports the results. The regressions confirm the general decrease in social value
orientation but show little significant differences between the Osdg benchmark and, respectively,
the market-treatments and trader roles. The exception are buyers in the market-strangers treatment who decrease their social value organization significantly more than participants in Osdg
(specification (3), p = 0.045). However, there are interesting differences between buyers and
sellers. When disregarding the market treatment we see that buyers decrease their social value
orientation significantly more than sellers (specification (2), p = 0.0217). Using specification (3),
Table 11 shows that this is mainly due to the difference in market-strangers (p < 0.001). Buyers
decrease their social value orientation more than sellers also in market-partners, but the decrease
is statistically not significant (p = 0.8662).
We summarize the effect of market experience on social value orientation in the next result.
Result 9. Social value orientation is generally decreasing over time. Overall, market-losers
(buyers) decrease their social value orientation more than market-winners (sellers). This effect
is significant only for market losers who interacted in market-strangers before.
7
Summary and conclusions
In this paper we explored whether competing in a market under favorable and unfavorable
conditions affects the efficiency of interaction in a subsequent social dilemma game. We find
that efficiency in the social dilemma is lower for people who did interact with each other in the
market before than for people who interacted in different markets or for people without market
experience. This is true both for market-losers and market-winners. It is as if the fact of having
been competitors on the same market hampers later cooperation between people. In this sense
competitive market experience is harmful for efficient cooperation.
We also find a positive effect of competition on the efficiency of cooperation, but only when
market-winners of two separate markets have been paired. Thus it appears that having had
a positive experience on a market leads to more efficient outcomes when the competition was
not with people that are – in some sense – close to each other. This finding resonates with the
observations of Camerer and Fehr (2004) in small-scale societies. They find that exposure to
market interactions with strangers can have a positive effect on social behavior towards partners
33
Table 10: Regression: change in social value orientation
Specification
Earnings in SDG 1
Market-partners
Market-strangers
Buyer
Seller
Market-partners*Buyer
Market-partners*Seller
Market-strangers*Buyer
Market-strangers*Seller
Constant (Osdg)
R2
N
F(1,57)
p-value
dep. var.: change in social value orientation
(1)
(2)
(3)
0.039
0.027
0.034
(0.023)
(0.020)
(0.023)
3.101
(2.510)
-2.047
(2.180)
-1.742
(1.887)
2.483
(2.386)
2.740
(2.361)
3.386
(4.088)
-4.276∗∗
(2.088)
1.842
(2.714)
-21.489∗∗
-16.165∗
-19.491∗
(10.217)
(8.572)
(10.015)
0.027
0.023
0.041
368
368
368
MPM vs MSM Seller vs Buyer
all comp. sig.
3.18
5.57
except
0.0800
0.0217
MSM*Seller vs MPM*Seller
Note: ∗∗∗ , (∗∗ ), [∗ ] indicated significance on 1, 5, and percent level, respectively. Standard errors corrected for dependence of observations for
58 clusters.
Table 11: Change in social value orientation: comparison of market roles
F(1,57)
p-value
F(1,57)
p-value
F(1,57)
p-value
MP*buyer=MP*seller
0.03
0.8662
MP*buyer=MS*buyer
7.02
0.0104
MP*buyer=MS*seller
0.08
0.7795
34
MS*buyer=MS*seller
12.51
0.0008
MP*seller=MS*seller
0.11
0.7431
MP*seller=MS*buyer
3.23
0.0776
from the own society.
We also find effects of market participation on subjective well-being and social value orientation. Our results show that for market-losers subjective well-being and social value orientation
both worsen compared to market-winners. In addition, regarding social value orientation marketstrangers worsen with respect to market-partners. A possible interpretation is that the higher
social distance in market-strangers has a more negative impact on social value orientation.
How can we reconcile this fact with the result that market-strangers interact more efficiently
than market-partners in the SDG environment? It turns out that lower satisfaction and a worse
social value orientation disposition towards others one has not interacted with in the market
does not impede more efficient cooperation. This suggests that, although in the case of marketstrangers participants are more negatively affected by the market-experience, they nevertheless
are more able to separate their experience in the market from the subsequent cooperation task.
Overall, our findings are in line with the view of Bowles (1998) presented in the introduction
but also refine it. Market experience does affect cooperation but it is neither market experience
per se nor being on the long or short side on the market that is adverse to efficient cooperation.
It is the combination of having competed with each other that makes subsequent cooperation
difficult, irrespective of having been a market-loser or market-winner.
35
References
Andreoni, J. (1988). Why free ride? Strategies and learning in public goods experiments. Journal
of Public Economics, 37:291304.
Bartling, B., Weber, R. A., and Yao, L. (2015). Do markets erode social responsibility? Quarterly
Journal of Economics, 130(1):219–266.
Bauernschuster, S., Falck, O., and Große, N. (2013). When trustors compete for the favour of a
trustee–a laboratory experiment. Journal of Economic Psychology, 34:133–147.
Bolton, G. and Ockenfels, A. (2000). ERC: A theory of equity, reciprocity, and competition.
American Economic Review, 90(1):166–193.
Bowles, S. (1998). Endogenous preferences: the cultural consequences of markets and other
economic institutions. Journal of Economic Literature.
Bowles, S. and Polania-Reyes, S. (2012). Economic incentives and social preferences: Substitutes
or complements? Journal of Economic Literature, 50(2):368–425.
Brandts, J., Riedl, A., and Winden, F. v. (2009). Competitive rivalry, social disposition, and
subjective well-being: An experiment. Journal of Public Economics, 93(11/12):1158–1167.
Buser, T. and Dreber, A. (2013). The flipside of comparative payment schemes. Tinbergen
Institute Discussion Paper, TI2013-190/I.
Camerer, C. F. and Fehr, E. (2004). Measuring social norms and preferences using experimental
games: A guide for social scientists. In Henrich, J., Boyd, R., Bowles, S., Camerer, C.,
Fehr, E., and Gintis, H., editors, Foundations of Human Sociality: Economic Experiments
and Ethnographic Evidence from Fifteen Small-Scale Societies, chapter 3, pages 55–95. Oxford
University Press, Oxford, UK.
Carpenter, J. and Seki, E. (2006). Competitive work environments and social preferences: Field
experimental evidence from a japanese fishing community. Contributions to Economic Analysis
& Policy, BE Press, 5(2). Article 2.
Charness, G. and Rabin, M. (2002). Understanding social preferences with simple tests. Quarterly
Journal of Economics, 117:817–869.
Chaudhuri, A. (2011). Sustaining cooperation in laboratory public goods experiments: a selective
survey of the literature. Experimental Economics, 14:47–83.
Chen, D. L. (2011). Markets and morality: How does competition affect moral judgement?
Working Paper.
Croson, R. T. A. (1996). Partners and strangers revisited. Economics Letters, 53:2532.
36
Davis, D. D. and Holt, C. A. (1993). Experimental economics, chapter 3. Double-auction markets.
Princeton University Press.
Engels, F. (1987/1845). The condition of the working class in England. Penguin UK. First
published 1845, Germany.
Falk, A. and Heckman, J. J. (2009). Lab experiments are a major source of knowledge in the
social sciences. Science, 326:535–538.
Falk, A. and Szech, N. (2013). Morals and markets. Science, 340:707–711.
Fehr, E. and Hoff, K. (2011). Tastes, castes and culture: The influence of society on preferences.
Economic Journal, 112(556):396–412.
Fehr, E. and Schmidt, K. (1999). A theory of fairness, competition, and cooperation. Quarterly
Journal of Economics, 114:817–868.
Fischbacher, U. (2007). z-tree: Zurich toolbox for ready-made economic experiments. Experimental Economics, 10(2):171–178.
Fourcade, M. and Healy, K. (2007). Moral views of market society. Annual Review of Sociology,
33.
Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., Gintis, H., and McElreath, R. (2001).
In search of homo economicus: Behavioral experiments in 15 small-scale societies. American
Economic Review, 91(2):72–78.
Herz, H. and Taubinsky, D. (2013). Market experience is a reference point in judgments of
fairness. Working Paper. Available at SSRN 2297773.
Holt, C. A., Langan, L., and Villamil, A. (1986). Market power in oral double auctions. Economic
Inquiry, 24:107–123.
Huck, S., Lünser, G. K., and Tyran, J.-R. (2012). Competition fosters trust. Games and
Economic Behavior, 76(1):195–209.
Kagel, J. H. and Roth, A. E., editors (2012). The Handbook of Experimental Economics, volume 2.
Princeton University Press, Princeton, New Jersey. forthcoming.
Lang, P. J. (1980). Behavioral treatment and bio-behavioral assessments: Computer applications.
In Sidowsky, J., Johnson, J., and Williams, T., editors, Technology in Mental Health Care
Delivery Systems. NJ Ablex, Norwood.
Liebrand, W. (1984). The effect of social motives, communication and group sizes on behaviour
in an N-person multi-stage mixed motive game.
European Journal of Social Psychology,
14:239264.
Mas-Colell, A., Whinston, M. D., and Green, J. R. (1995). Microeconomic Theory. Oxford
37
University Press, Oxford, UK.
Montesquieu, C. B. d. S. (1748). The Spirit of Laws. Crowder, Wark, and Payne, 1777. Originally
published anonymously. Trans. Thomas Nugent (1750).
Murphy, R. O., Ackermann, K. A., and Handgraaf, M. J. J. (2011). Measuring social value
orientation. Judgment and Decision Makin, 6(8):771–781.
Noussair, C. and Tucker, S. (2013). A Collection of Surveys on Market Experiments. Surveys of
Recent Research in Economics. Wiley.
Sandel, M. J. (2012). What Money Can’t Buy: The Moral Limits of Markets. Farrar, Straus and
Giroux, New York.
Sandel, M. J. (2013). Market reasoning as moral reasoning: why economists should re-engage
with political philosophy. The Journal of Economic Perspectives, pages 121–140.
Smith, V. (1962). An experimental study of competitive market behavior. Journal of Political
Economy, 70:111–137.
Smith, V. (1976). Experimental economics: induced value theory. American Economic Review,
66:274–279.
Smith, V. (1998). The two faces of adam smith. Southern Economic Journal, 65(1):1–19.
Sonnemans, J. (1991). Structure and Determinants of Emotional Intensity. PhD thesis, University of Amsterdam.
Sonnemans, J., van Dijk, F., and van Winden, F. (2006). On the dynamics of social ties structures
in groups. Journal of Economic Psychology, 27:187204.
Standing, G. (2011). The Precariat; The Dangerous New Class. Bloomsbury Academic.
Van Lange, P. A. M. (1999). The pursuit of joint outcomes and equality in outcomes: An
integrative model of social value orientation. Journal of Personality and Social Psychology,
77(2):337–349.
van Winden, F. (2012). Affective social ties - Missing link in governance theory. Rationality,
Morals and Markets, 3(108-122).
38
A
Additional Statistical Results
A.1
Nonparametric Statistics
Incomplete.
For initial subjective well-being no significant difference between treatments found (KruskalWallis test, p = 0.3494, 2-sided).
For final subjective well-being test on differences across treatments without distinguishing
between sellers and buyers (Kruskal-Wallis test, p = 0.1096, 2-sided; based on independent observations). Pair-wise MW tests: osdg>mpm (p = 0.0643), osdg=msm (p = 0.1525), msm=mpm
(p = 0.5194)
For change in subjective well-being test on differences across treatments without distinguishing between sellers and buyers (Kruskal-Wallis test, p = 0.2798, 2-sided; based on independent
observations).
For initial social value orientation marginally significant difference between treatments found
(Kruskal-Wallis test, p = 0.0623, 2-sided). Significance disappears after excluding two extreme
observations (Kruskal-Wallis test, p = 0.1044, 2-sided).
For final social value orientation significant difference between treatments found (KruskalWallis test, p = 0.0703, 2-sided; based on independent observations).
Pair-wise MW-tests: osdg>mpm (p = 0.0336), osdg=msm (p = 0.1966), msm=mpm (p =
0.3036)
For change in social value orientation no significant difference between treatments found
(Kruskal-Wallis test, p = 0.5012, 2-sided; based on independent observations).
39
A.2
Additional Regression Analysis
Incomplete?
Table 12: Contributions in the social dilemma game: comparing trader matchings within market
treatments (all periods)
Specification
buyer-buyer
seller-seller
seller-buyer
Constant (Osdg)
R2
N
F
p-value
F
p-value
F
p-value
dep. var.: contribution to public good
(1)
(2)
Market-partners
Market-strangers
-9.678∗∗
-0.970
(3.778)
(3.135)
-8.319∗
5.899∗
(4.264)
(3.297)
-0.000
1.557
(4.760)
(3.442)
29.375∗∗∗
29.375∗∗∗
(2.646)
(2.647)
0.058
0.018
3168
4608
buyer-buyer vs seller-seller
0.09
5.77
0.7624
0.0207
buyer-buyer vs seller-buyer
4.32
0.62
0.0433
0.4356
seller-buyer vs seller-seller
2.58
2.30
0.1150
0.1365
Note: ∗∗∗ , (∗∗ ), [∗ ] indicated significance on 1, 5, and 10 percent level,
respectively. Standard errors corrected for dependence of observations for
respectively 46 and 44 clusters in specification (1) and (2). F -test is F1,45
(F1,43 ) for market-partners (market-strangers).
40
Table 13: Contributions in the social dilemma game: comparing trader matchings within market
treatments (periods 1-6 & periods 7-18)
Specification
buyer-buyer
seller-seller
seller-buyer
Constant (Osdg)
R2
N
F
p-value
F
p-value
F
p-value
Specification
buyer-buyer
seller-seller
seller-buyer
Constant (Osdg)
R2
N
F
p-value
F
p-value
F
p-value
Note:
∗∗∗
Periods 1-6
dep. var.: contribution to public good
(1)
(2)
Market-partners
Market-strangers
-6.721∗
-0.786
(3.345)
(3.071)
-6.683
7.036∗∗
(4.150)
(3.327)
-1.576
3.265
(4.806)
(3.316)
28.451∗∗∗
28.451∗∗∗
(2.597)
(2.597)
0.030
0.020
1056
1536
buyer-buyer vs seller-seller
0.00
4.42
0.9919
0.0414
buyer-buyer vs seller-buyer
1.68
0.63
0.2020
0.4316
seller-buyer vs seller-seller
0.98
2.32
0.3272
0.1349
Periods 7-18
dep. var.: contribution to public good
(1)
(2)
Market-partners
Market-strangers
-11.156∗∗
-1.848
(4.183)
(3.390)
-9.138∗
5.331
(4.692)
(3.488)
0.788
0.703
(5.081)
(3.792)
29.837∗∗∗
28.837∗∗∗
(2.823)
(2.824)
0.075
0.018
2112
3072
buyer-buyer vs seller-seller
0.15
5.52
0.7025
0.0234
buyer-buyer vs seller-buyer
5.35
0.54
0.0253
0.4656
41
seller-buyer vs seller-seller
2.87
1.85
0.0970
0.1810
, (∗∗ ), [∗ ] indicated significance on 1, 5, and 10 percent level,
Table 14: Contributions in the social dilemma game: comparing trader matchings within market
treatments and earnings matchings within Osdg-Income (all periods)
Specification
buyer-buyer
seller-seller
seller-buyer
low-low
high-high
high-low
Constant (Osdg)
R2
N
F
p-value
F
p-value
F
p-value
F
p-value
F
p-value
F
p-value
dep. var.: contribution to public good
(1)
(2)
Market-partners
Market-strangers
Osdg-Income
Osdg-Income
∗∗
-9.678
-0.970
(3.760)
(3.119)
-8.319∗
5.899∗
(4.244)
(3.281)
-0.000
1.557
(4.738)
(3.425)
4.065
3.940
(4.399)
(4.117)
3.139
-0.347
(4.791)
(4.294)
1.005
3.940
(5.178)
(4.117)
29.375∗∗∗
29.375∗∗∗
(2.634)
(2.634)
0.058
0.016
4464
5904
high-high vs high-low
0.13
0.01
0.7226
0.9371
high-high vs low-low
0.03
0.85
0.8626
0.3581
high-low vs low-low
0.29
0.52
0.5917
0.4743
high-high vs seller-seller
4.85
2.55
0.0305
0.1146
low-low vs buyer-buyer
9.63
1.88
0.0026
0.1146
high-low vs seller-buyer
0.03
0.09
0.8663
0.7626
Note: ∗∗∗ , (∗∗ ), [∗ ] indicated significance on 1, 5, and 10 percent level,
respectively. Standard errors corrected for dependence of observations for
respectively 46 and 44 clusters in specification (1) and (2). F -test is F1,45
(F1,43 ) for market-partners (market-strangers). low-low, high-high, and
high-low stands for 2672-2672, 340-340, and 2672-340 in market-partners
and 2656-2656, 346-346, and 2656-346 in market-strangers.
42
A.3
Additional Figures
50
50
45
45
40
40
35
35
mean contribution
mean contribution
Incomplete?
30
25
20
15
30
25
20
15
10
10
5
5
0
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
mpm: buyer−buyer
mpm: seller−buyer
1
mpm: seller−seller
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
msm: buyer−buyer
msm: seller−buyer
(a) Market-partners
msm: seller−seller
(b) Market-strangers
Figure 7: Average contributions over rounds in the different matchings by market treatments
43
50
45
45
40
40
35
35
mean contribution
mean contribution
50
30
25
20
15
30
25
20
15
10
10
5
5
0
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
Market−partners
OSDG−INCOME: 340−340
1
Market−strangers
OSDG−INCOME: 346−346
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
Market−partners
OSDG−INCOME: 2656−2656
(a) Buyer-buyer & 340-340 & 346-346
Market−strangers
OSDG−INCOME: 2672−2672
(b) Seller-seller & 2656-2656 & 2672-2672
50
45
mean contribution
40
35
30
25
20
15
10
5
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14 15 16 17 18
period
Market−partners
OSDG−INCOME: 340_2672
Market−strangers
OSDG−INCOME: 346_2656
(c) Seller-buyer & 340-2672 & 346-2656
Figure 8: Average contributions over rounds in the both market treatments and in OsdgCONTROL-Income by different trader and corresponding endowment matchings
44
A.4
Experiment Instructions
Incomplete!
45